CARS Preview 2018

Project by RSIP Vision 25 CARS 2018 Preview RSIP Vision, the publisher of this CARS Preview, is actually not a publishing house. It’s a software house that makes state-of-the-art research in partnership with major medical centers. Here is a very recent R&D project, completed by RSIP Vision in cooperation with Tufts. Findings were presented at the annual meeting of ARVO, the Association for Research in Vision and Ophthalmology. Introduction: In Vivo Confocal Microscopy (IVCM) is a non-invasive imaging technique enables to acquire coronal images throughout the cornea at a cellular level. Through these images it is possible to assess the Dendritic Cells (DCs) density, which is strongly correlated with Dry Eye Disease (DED), one of the most frequent ophthalmic disorders . Therefore, this tool provides a non-invasive biomarker for corneal inflammation. Our study proves that automated Convolutional Neural Networks (CNN) detect and quantify DCs, providing an accurate estimation of ocular surface inflammation and significant benefits for diagnosis and treatment. Method: A ‘U-Net’ based fully convolutional neural network (CNN) architecture was developed to analyze the 2140 IVCM images from our database: images demonstrative of healthy epithelium, subbasal nerve plexus, stroma, and endothelium. Training and validation datasets were selected randomly from the database. Deep Learning Convolutional Neural Network for the Classification and Segmentation of In Vivo Confocal Microscopy Images